16

I am currently working with Salvadoran Nawat, an endangered language that has never had a standardized orthography due to being primarily oral. As part of the revitalization process, we need to standardize an orthography for this language. However, we are not sure if we should represent phonemes or phones in the orthography. For instance, in cases where two consecutive /k/ phonemes occur as in /kk/, the pronunciation in Nawat actually entails aspiration of the first k, resulting in [hk]. Consequently, a word like /nutsakka/ is articulated as [nutsahka], should we write it as nutzakka or nutzahka? Similarly, when /n/ precedes /p/, as in /senpa/, the articulation in Nawat is [sempa]. Should we write senpa or sempa in such instances?

In the context of revitalizing endangered languages that have primarily been oral, should orthographies represent phonemes or phones? How do we address the issue of phonemic vs. phonetic orthographical representation in cases where there are differences in pronunciation such as the example of /kk/ becoming [hk] in Salvadoran Nawat?

Sir Cornflakes
  • 30,154
  • 3
  • 65
  • 128
Sigfredo Olmedo
  • 161
  • 4
  • 10
    My answer would normally be "it depends, you need to provide more detail", but you've laid out your use case pretty well. The short answer is that writing phonemes is more useful to native speakers and writing phones is more useful to outsiders, but I'll try to write up a more comprehensive explanation tomorrow. – Draconis Mar 27 '23 at 04:55
  • If those sound changes happen in all the dialects of that language, what arguments are there in favor of the phonemic spelling? Writing the way one hears it is the most natural way of writing, isn't it? – Yellow Sky Mar 27 '23 at 04:57
  • 6
    @YellowSky but native speakers tend not to hear distinct allophones, instead hearing a unified phoneme. If writing the way one hears is the most natural way, the question is whose hearing should we write according to – Tristan Mar 27 '23 at 08:46
  • @Tristan — My whole life's experience says the opposite: native speakers tend to hear phones, not unified phonemes. That's the main issue in teaching the Russian children to write Russian: they tend to write phones while the Russian spelling is phonological. Before mastering the spelling rules application children reflect in writing the word-final stops devoicing, assimilation in cons. clusters and the unstressed vowels reduction, which is never shown in the standard spelling since it writes phonemes, not the phones. So what makes you think native speakers tend not to hear distinct allophones? – Yellow Sky Mar 27 '23 at 11:27
  • 8
    @YellowSky Are those actually the same phonemes, though, and not just the same morphophonemes? Anything termed ‘word-final’ anything, at least, must by definition involve morphemes, not just phonemes. Native English speakers, including children, will virtually always hear the /l/ in plaque and black or the /e/ in bed and bet as the same sound, not as separate sounds. Similarly, (adult) Spanish speakers will swear up and down that the two ’s in un dedo are pronounced the same, despite them being acoustically quite obviously distinct. – Janus Bahs Jacquet Mar 27 '23 at 13:19
  • 1
    @Sigfredo Even [nutzahka] would be quite unusual, phonetically: [t] is an unvoiced stop, while [z] is a voiced sibilant fricative, and mixed-voicing clusters of this type are typologically rare. Is it actually pronounced with an unvoiced [t] and a voiced [z], or is it really pronounced [ts]? – Janus Bahs Jacquet Mar 27 '23 at 13:22
  • 3
    @YellowSky children are still forming their model of the language's phonology so aren't a good indication of how adult native speakers perceive things (completely illiterate adults would be ideal informants, as they'd have less interference from spelling, but in standard languages there aren't many of those without other complicating factors) – Tristan Mar 27 '23 at 13:32
  • 5
    @YellowSky Regardless, English-speakers generally cannot hear the difference between the l's in lick & kill or the t's in stop & top (and certainly don't hear the t in store as the same as the d in door, despite them being phonetically identical in most varieties) – Tristan Mar 27 '23 at 13:32
  • 1
    Literate English speakers are hard cases for hearing any sounds. It's drummed into them for 8 or more years that graphemes make the sounds, and if you hear anything else, you're not only wrong but failing as an English speaker. – jlawler Mar 27 '23 at 15:07
  • @Tristan — The difference between those allophones of l's and t's is so minute that there's even no way in English to mark that difference in writing. But in Russian when зуб (zub) /zub/ “tooth” is pronounced as [zup] by everyone and there's a special letter <п> for the sound [p], doesn't it seem natural to write it as it's heard, as зуп (zup)? That's what I mean. And no, when children are in the age they start learning to write they are not still forming their model of the language's phonology, they've already formed it. – Yellow Sky Mar 27 '23 at 15:08
  • @YellowSky That's the origin of English "eye dialect" like gotcha and hafta. They're accurate misspellings. In the various Salishan languages I've got dictionaries of (many more coming out recently, as the Nations get their language revitalization schemes moving), every possible kind of scheme is being used, because every group makes their own decisions about what they want. So there's no "should", unless one is speaking for the group, at their behest. They should make all the decisions, since they'll hafta live with them. – jlawler Mar 27 '23 at 15:15
  • @Tristan — Writing phonemes implies being able to perform phonological and morphological analysis of the words which is a skill requiring additional training while writing as it's heard needs no other skill than just knowing the shape and the sound value of the letters, so what's easier? There's the Belarusian language, very close to Russian, but Belarusian spelling is practically purely phonetic, so when Russians see a Belarusian text for the first time in their lives they are shocked that a language can function well when it is written exactly as it's heard, as a child would write Russian. – Yellow Sky Mar 27 '23 at 15:18
  • @JanusBahsJacquet you're right, I actually messed up, it's [nutsahka] – Sigfredo Olmedo Mar 27 '23 at 15:35
  • 1
    @YellowSky The fact that there’s no way in English orthography to mark the difference between the two l’s is rather a straw man: English is a highly etymologising morphophemic orthography, and even when an allophone exists as the default form of a separate phoneme, that’s often not reflected in writing (e.g., devoicing of -s and -ed), so it’s fully expected that allophones that don’t overlap with another phoneme are not distinctly writable. For the two t’s, though, there is a way to write the difference if you wanted to: use for the unaspirated variant. – Janus Bahs Jacquet Mar 27 '23 at 17:13
  • 1
    @YellowSky And at any rate, your initial claim was that speakers generally hear phones, not phonemes. The fact remains that English speakers hear the stops in store and tore as identical, but the ones in store and door as different, which direct contradicts that claim. If speakers instinctively heard phones and not phonemes, they would hear the stops in store and door as identical (since they are, phonetically) and the ones in store and tore as different (since they are, phonetically). That is one valid reason for spelling in phonemes more than surface phones. – Janus Bahs Jacquet Mar 27 '23 at 17:18
  • @JanusBahsJacquet - For some reasons it always seemed to me that [d] in door is voiced while [t] in store is voiceless, and now you say they are identical phonetically, so what did I miss? What's your definition of “identical”? Perhaps I use this word to mean something different, what do you think? – Yellow Sky Mar 27 '23 at 17:54
  • 3
    @YellowSky To the majority of speakers, the d in door is unvoiced (or mostly unvoiced) in most contexts. Voicing is only even remotely consistent postvocalically. To be more precise, I should have said “since they generally are”. – Janus Bahs Jacquet Mar 27 '23 at 18:10
  • 2
    @YellowSky In actual fact, they aren't identical. There are all sorts of other differences between the two, but most of these are very weak cues, if even audible, for listeners. However, they are both likely to be unvoiced when not preceded by a voiced phone. Why does your language brain tell you that the [d] in door is voiced when not preceded by a vowel? It's because the [d]'s not aspirated. In other words it is the devoicing of the vowel that gives you the cue telling you it's a [d] not a [t]. – Araucaria - him Mar 29 '23 at 13:18
  • @YellowSky Sorry, the [d] and the [t] are the wrong way round in the last sentence above. – Araucaria - him Mar 30 '23 at 09:44
  • 1
    @YellowSky My native language is Croatian (also Slavic lang.) and I also speak Russian. I am a bit shocked to hear that you actually pronounce зуп, not зуб. In Croatian we strictly pronounce "(б / b)". I checked on Forvo, somewhere I hear /zup/, somewhere /zub/. I am curious, how would you write other cases where there is clearly /b/ pronounced, for example зубов? Or plural зубы? I think it would be rather confusing if sometimes you write п and sometimes б for the same word. – dosvarog Apr 01 '23 at 14:15

6 Answers6

22

Consult the speech community.

The orthography must fit the needs of the speech community, they are the primary users of it. When the speech community wants a phonetic representation (helping essentially L2 learners), do it. When the speech community prefers a phonemic representation (benefitting L1 speakers and imposing additional effort on L2 speakers), do that.

It is the speech community that has the final vote on a practical orthography.

Sir Cornflakes
  • 30,154
  • 3
  • 65
  • 128
6

It does not per se matter whether you write using narrow vs broad transcriptions. The following desiderata should guide your choices. Number 1 is, do speakers like your choice over the alternative. That alone is very difficult to decide factually. In aid of figuring out what possibilities people will like, the primary considerations are ease of use, and representational adequacy (not creating massive ambiguity). To the extent that there are phonological alternations in the language, a more-phonetic spelling means that it becomes harder for a user to say what the root for "tamale" is in the language. Ideally, each morpheme has a single invariant spelling. However, if the language has a very active phonology where affixation produces substantial alternations, and when the underlying form can't be easily discerned from the citation form, it may be necessary to introduce orthographic alternations which parallel the phonological alternations.

Let's assume this dialects has autonomous /ŋ/ in <tenhat> ([ŋ]) 'river bank'. An adequate spelling system would have some means of indicating this. This means that /ŋ/ is a phoneme, and in a purely phonemic writing system you would therefore write teŋkal as <tenhkal>, but I expect that speakers will think that this clutters the writing system, in being unnecessarily faithful to pronunciation. The general principle is that if there is a simple rule that predicts the pronunciation from spelling, you should exploit that rule.

Another consideration is whether the phonemic value of some sound is highly obvious to anyone with minimal education. You claim that there is /kk/ but it is pronounced [hk] – what is your argument that it is indeed /kk/ and not /hk/? Alternations might make it obvious, e.g. /tek-ka/ → [tehka] but /tek-ta/ → [tekta]. An argument like "we don't have to have underlying geminates" is not a very useful argument, even professional linguists recognize that that is not strong motivation for positing a divergence from pronunciation in underlying forms (a concept that is close to useless to a person learning how to spell).

user6726
  • 83,066
  • 4
  • 63
  • 181
4

The answer of user6726 raises some very good points, but really the correct answer is that you should get in touch with someone (ideally more than one person) who has faced this problem before, instead of philosophizing ex nihilo about what might be the better choice.

You are facing an unusual and highly non-trivial task that relatively few people have done before and that your experience has probably not prepared you for. (Presumably you wouldn't be asking here if you've already done this before.) Do you want to repeat the mistakes that people made in the past when devising new orthographic systems, or do you want to learn from their experience? Note that if you make a mistake, it can be quite hard to fix (example).

If you were planning an Arctic expedition, would you philosophize about the merits and demerits of one might possibly want to take with them or which means of transportation to choose with strangers on the internet (no matter how knowledgeable), or would you ask someone who has actually been on an Arctic expedition on their own?

Let me also add that the idea that "the speech community" will make decision for you may very well be false hope (unless the speech community happens to have a strong opinion on every design choice that you face). Consult them, for sure, but they have never face this task before any more than you have. "The speech community" is not an oracle for you to consult and abide by its declarations, it's a sounding board that might agree or disagree with certain points but will presumably also be open to persuasion. Of course you cannot force your preferences on them, but it's not your role an an expert to blindly follow whatever the speech community tells you.

Pilcrow
  • 156
  • 4
  • 3
    Oh my, where would one find people who’d had experience of dealing with newly invented orthographies for natural languages I wonder? – Araucaria - him Mar 28 '23 at 13:34
  • 3
    It’s not the designers you need to speak to it’s the users who are left to live with the system long after the designers have disappeared. And it is eminently sensible to ask speakers what they think a particular sound in a word is or represents. It makes no difference whatsoever if they’ve ever designed an orthographic system for a natural language before. Why should it? – Araucaria - him Mar 28 '23 at 13:39
  • @Araucaria-him Interaction through Linguistics SE based on an isolated example of a single word (with people who, by the looks of it, for the most part have not, contrary to your insinuation, had a hand in designing any orthographic system) is hardly a substitute for an in-depth conversation with someone with hands-on experience. Secondly, you should of course speak to designers who kept in touch with the speech community. I imagine that if a designer is involved enough with the speech community to come up with an orthography, then they perhaps in seeing how it actually works in practice. – Pilcrow Mar 28 '23 at 19:23
  • 1
    @Araucaria-him "And it is eminently sensible to ask speakers what they think a particular sound in a word is or represents." Yes, of course it is. But it seems bizarre to deny that people who have hands-on experience with designing orthographic systems can share useful experience that ordinary speakers cannot. And if you're not doing that, then I don't know what point you are trying to make here. – Pilcrow Mar 28 '23 at 19:24
  • 2
    They don't need to have had experience of designing the systems; they need to be familiar with the main issues in graphisation or have had experience of other natural languages which have had orthographies put together. Some members here may have some knowledge of several different recently invented orthographies (and their consequences). This is amply sufficient for professional linguists(to be able to have views on the subject and to offer advice, especially if their area of expertise includes endangered languages, for instance, and in many other cases too. – Araucaria - him Mar 29 '23 at 09:40
  • @Araucaria-him Yes, sure, you're absolutely right that it isn't necessary to have personally had a hand in inventing an orthography, it is entirely sufficient to have other kinds of experience with recently invented orthographies. I also don't doubt the collective expertise of the Linguistics SE community. On the other hand, I think my main point still stands, namely that the OP should arrange to have an in-depth conversation with someone like that, instead of getting very general feedback based on an isolated example of a single word at Linguistics SE. – Pilcrow Mar 29 '23 at 10:44
3

I'm a conlanger, rather than a professional linguist, but I use a unique letter for each phoneme, but whenever that sound occurs, even as an allophone, it is written that way.

In other words, if /m/ is a phoneme as well as an allophone of /n/ before labial consonants, [sempa] should be written sempa, but if /m/ is not an independent phoneme, senpa makes more sense.

Of course, this is also a matter of personal preference. But a more phonemic writing system would be easier for the native speakers to acquire, and therefore more useful for the revitalization.

nearsighted
  • 395
  • 1
  • 11
  • 3
    That's the normal phonemic orthographic principle. As Pike subtitled his book Phonemics: A technique for reducing languages to writing. University of Michigan Publications Linguistics 3. Ann Arbor. 1947. And the reduction was to alphabetic writing, by phoneme. – jlawler Mar 27 '23 at 15:10
1

Also, in addition the phonetics, I suggest the principle to keep the orthography consistent with the grammar.

As instance, in German (the language reform of 1998 simplified the grammar), Schifffahrt has 3 "f" because it is compound of Schiff+fahrt, while the old word Schiffahrt had to loose one "f" [for typographical reasons, as suggested in a comment below].

The same principle of simplification is for the usage of ss or ß:

  • New rule: Use ss, if it is preceded by a short vowel, use ß otherwise.
  • Old rule: Use ss, if it is preceded by a short vowel and the second s starts a new syllable, use ß otherwise [as suggested in a comment below].
blue_lama
  • 37
  • 6
  • 2
    Old German orthography used double consonants where triple ones are expected mainly for typographical reasons, not for phonetical ones. – Sir Cornflakes Mar 29 '23 at 08:58
  • Thank you, I learnt something. – blue_lama Mar 29 '23 at 14:31
  • 1
    New rule: Use "ss", if it is preceded by a short vowel, use "ß" otherwise. Old rule: Use "ss", if it is preceded by a short vowel and the second "s" starts a new syllable, use "ß" otherwise. – Uwe Apr 04 '23 at 13:56
1

I'm not convinced that an accurate phonetic transcription would help L2 learners (let alone L1 ones). Co-articulation is both universal and not really consciously perceived (see also above comment by Tristan). Most of the time, when isolating bits of perfectly normal speech, you are surprised to hear how much centered the vowels are compared to their targets, or how sloppy the realization of a stop consonant is. So my hunch would be to stick to phonemes and forget the phones, that are very context-dependent.

Anyway, what you're doing is not there to stay indefinitely, spelling (and language) will evolve after you. Thompson finally got a 'p' because it can be (illusorily) heard in the realization of Thomson. In this case the perceived phone imposed its presence in a language whose spelling is not phonetic... (but this only happened because the meaning "son of Thom" was progressively lost).

Pedrok
  • 11
  • 2